- 17 Apr, 2021 4 commits
-
-
Srikar Dronamraju authored
On systems with large CPUs per node, even with the filtered matching of related CPUs, there can be large number of calls to cpu_to_chip_id for the same CPU. For example with 4096 vCPU, 1 node QEMU configuration, with 4 threads per core, system could be see upto 1024 calls to cpu_to_chip_id() for the same CPU. On a given system, cpu_to_chip_id() for a given CPU would always return the same. Hence cache the result in a lookup table for use in subsequent calls. Since all CPUs sharing the same core will belong to the same chip, the lookup_table has an entry for one CPU per core. chip_id_lookup_table is not being freed and would be used on subsequent CPU online post CPU offline. Reported-by:
Daniel Henrique Barboza <danielhb413@gmail.com> Suggested-by:
Michael Ellerman <mpe@ellerman.id.au> Signed-off-by:
Srikar Dronamraju <srikar@linux.vnet.ibm.com> Tested-by:
-
Srikar Dronamraju authored
Now that cpu_core_mask has been reintroduced, lets revert commit 4bce5459 ("powerpc/topology: Update topology_core_cpumask") Post this commit, lscpu should reflect topologies as requested by a user when a QEMU instance is launched with NUMA spanning multiple sockets. Reported-by:
-
Srikar Dronamraju authored
Daniel reported that with Commit 4ca234a9 ("powerpc/smp: Stop updating cpu_core_mask") QEMU was unable to set single NUMA node SMP topologies such as: -smp 8,maxcpus=8,cores=2,threads=2,sockets=2 i.e he expected 2 sockets in one NUMA node. The above commit helped to reduce boot time on Large Systems for example 4096 vCPU single socket QEMU instance. PAPR is silent on having more than one socket within a NUMA node. cpu_core_mask and cpu_cpu_mask for any CPU would be same unless the number of sockets is different from the number of NUMA nodes. One option is to reintroduce cpu_core_mask but use a slightly different method to arrive at the cpu_core_mask. Previously each CPU's chip-id would be compared with all other CPU's chip-id to verify if both the CPUs were related at the chip level. Now if a CPU 'A' is found related / (unrelated) to another CPU 'B', all the thread siblings of 'A' and thread siblings of 'B' are automatically marked as related / (unrelated). Also if a platform doesn't support ibm,chip-id property, i.e its cpu_to_chip_id returns -1, cpu_core_map holds a copy of cpu_cpu_mask(). Fixes: 4ca234a9 ("powerpc/smp: Stop updating cpu_core_mask") Reported-by:
-
Cédric Le Goater authored
The 'chip_id' field of the XIVE CPU structure is used to choose a target for a source located on the same chip. For that, the XIVE driver queries the chip identifier from the "ibm,chip-id" property and compares it to a 'src_chip' field identifying the chip of a source. This information is only available on the PowerNV platform, 'src_chip' being assigned to XIVE_INVALID_CHIP_ID under pSeries. The "ibm,chip-id" property is also not available on all platforms. It was first introduced on PowerNV and later, under QEMU for pSeries/KVM. However, the property is not part of PAPR and does not exist under pSeries/PowerVM. Assign 'chip_id' to XIVE_INVALID_CHIP_ID by default and let the PowerNV platform override the value with the "ibm,chip-id" property. Signed-off-by:
Cédric Le Goater <clg@kaod.org> Signed-off-by:
-
- 16 Apr, 2021 4 commits
-
-
Tyrel Datwyler authored
The pci_bus->bridge reference may no longer be valid after pci_bus_remove() resulting in passing a bad value to device_unregister() for the associated bridge device. Store the host_bridge reference in a separate variable prior to pci_bus_remove(). Fixes: 73400565 ("powerpc/pci: Reorder pci bus/bridge unregistration during PHB removal") Signed-off-by:
Tyrel Datwyler <tyreld@linux.ibm.com> Signed-off-by:
-
Michael Ellerman authored
Now that __fake_sleep is static, we get a warning about it being unused in some configurations: drivers/macintosh/via-pmu.c:190:12: warning: '__fake_sleep' defined but not used 190 | static int __fake_sleep; Move it inside the ifdef where it's used to avoid the warning. Fixes: 95d14392 ("macintosh/via-pmu: Make some symbols static") Reported-by:Stephen Rothwell <sfr@canb.auug.org.au> Signed-off-by:
-
Michael Ellerman authored
When I changed the rc variable to be long rather than int64_t I neglected to update the printk(), leading to a build break: arch/powerpc/platforms/pseries/papr_scm.c: In function 'papr_scm_pmem_flush': arch/powerpc/platforms/pseries/papr_scm.c:144:26: warning: format '%lld' expects argument of type 'long long int', but argument 3 has type 'long int' [-Wformat=] Fixes: 75b7c05e ("powerpc/papr_scm: Implement support for H_SCM_FLUSH hcall") Reported-by: -
Michael Ellerman authored
This is a pseries only driver, it should be built by default as part of pseries_defconfig to get some build coverage. Signed-off-by:
-
- 14 Apr, 2021 32 commits
-
-
Michael Ellerman authored
The lkp bot pointed out that with W=1 we get: arch/powerpc/mm/book3s64/radix_pgtable.c:183:6: error: no previous prototype for 'radix__change_memory_range' Which is really saying that it could be static, make it so. Reported-by:
kernel test robot <lkp@intel.com> Signed-off-by:
-
Christophe Leroy authored
This patch adds the necessary glue to provide time namespaces. Things are mainly copied from ARM64. __arch_get_timens_vdso_data() calculates timens vdso data position based on the vdso data position, knowing it is the next page in vvar. This avoids having to redo the mflr/bcl/mflr/mtlr dance to locate the page relative to running code position. Signed-off-by:
Christophe Leroy <christophe.leroy@csgroup.eu> Reviewed-by: Vincenzo Frascino <vincenzo.frascino@arm.com> # vDSO parts Acked-by:
Andrei Vagin <avagin@gmail.com> Signed-off-by:
-
Dmitry Safonov authored
Since commit 511157ab ("powerpc/vdso: Move vdso datapage up front") VVAR page is in front of the VDSO area. In result it breaks CRIU (Checkpoint Restore In Userspace) [1], where CRIU expects that "[vdso]" from /proc/../maps points at ELF/vdso image, rather than at VVAR data page. Laurent made a patch to keep CRIU working (by reading aux vector). But I think it still makes sence to separate two mappings into different VMAs. It will also make ppc64 less "special" for userspace and as a side-bonus will make VVAR page un-writable by debugger (which previously would COW page and can be unexpected). I opportunistically Cc stable on it: I understand that usually such stuff isn't a stable material, but that will allow us in CRIU have one workaround less that is needed just for one release (v5.11) on one platform (ppc64), which we otherwise have to maintain. I wouldn't go as far as to say that the commit 511157ab is ABI regression as no other userspace got broken, but I'd really appreciate if it gets backported to v5.11 after v5.12 is released, so as not to complicate already non-simple CRIU-vdso code. Thanks! [1]: https://github.com/checkpoint-restore/criu/issues/1417 Cc: stable@vger.kernel.org # v5.11 Signed-off-by:
Dmitry Safonov <dima@arista.com> Signed-off-by:
-
Christophe Leroy authored
For the same reason as commit e876f0b6 ("lib/vdso: Allow architectures to provide the vdso data pointer"), powerpc wants to avoid calculation of relative position to code. As the timens_vdso_data is next page to vdso_data, provide vdso_data pointer to __arch_get_timens_vdso_data() in order to ease the calculation on powerpc in following patches. Signed-off-by:
Thomas Gleixner <tglx@linutronix.de> Reviewed-by:
Vincenzo Frascino <vincenzo.frascino@arm.com> Acked-by:
-
Christophe Leroy authored
In the same spirit as commit c966533f ("lib/vdso: Mark do_hres() and do_coarse() as __always_inline"), mark do_hres_timens() and do_coarse_timens() __always_inline. The measurement below in on a non timens process, ie on the fastest path. On powerpc32, without the patch: clock-gettime-monotonic-raw: vdso: 1155 nsec/call clock-gettime-monotonic-coarse: vdso: 813 nsec/call clock-gettime-monotonic: vdso: 1076 nsec/call With the patch: clock-gettime-monotonic-raw: vdso: 1100 nsec/call clock-gettime-monotonic-coarse: vdso: 667 nsec/call clock-gettime-monotonic: vdso: 1025 nsec/call Signed-off-by:
-
Nicholas Piggin authored
Compact the trap flags down to use the low 4 bits of regs.trap. A few 64e interrupt trap numbers set bit 4. Although they tended to be trivial so it wasn't a real problem[1], it is not the right thing to do, and confusing. [*] E.g., 0x310 hypercall goes to unknown_exception, which prints regs->trap directly so 0x310 will appear fine, and only the syscall interrupt will test norestart, so it won't be confused by 0x310. Signed-off-by:Nicholas Piggin <npiggin@gmail.com> Signed-off-by:
-
Nicholas Piggin authored
All subarchitectures always save all GPRs to pt_regs interrupt frames now. Remove FULL_REGS and associated bits. Signed-off-by:
-
Nicholas Piggin authored
search_exception_tables + __bad_page_fault can be substituted with bad_page_fault, do_page_fault no longer needs to return a value to asm for any sub-architecture, and __bad_page_fault can be static. Signed-off-by:
-
Nicholas Piggin authored
With non-volatile registers saved on interrupt, bad_page_fault can now be called by do_page_fault. Signed-off-by:
-
Nicholas Piggin authored
With the new interrupt exit code, context tracking can be managed more precisely, so remove the last of the 64e workarounds and switch to the new context tracking code already used by 64s. Signed-off-by:
-
Nicholas Piggin authored
Use existing 64s interrupt entry wrapper code to reconcile irqs in C. Signed-off-by:
-
Nicholas Piggin authored
64e non-maskable interrupts save the state of the irq soft-mask in asm. This can be done in C in interrupt wrappers as 64s does. I haven't been able to test this with qemu because it doesn't seem to cause FSL bookE WDT interrupts. This makes WatchdogException an NMI interrupt, which affects 32-bit as well (okay, or create a new handler?) Signed-off-by:
-
Nicholas Piggin authored
Update the new C and asm interrupt return code to account for 64e specifics, switch over to use it. The now-unused old ret_from_except code, that was moved to 64e after the 64s conversion, is removed. Signed-off-by:
-
Nicholas Piggin authored
This makes adjustments to 64-bit asm and common C interrupt return code to be usable by the 64e subarchitecture. Signed-off-by:
-
Nicholas Piggin authored
In order to use the C interrupt return, nvgprs must always be saved. Signed-off-by:
-
Nicholas Piggin authored
user_exit_irqoff() -> __context_tracking_exit -> vtime_user_exit warns in __seqprop_assert due to lockdep thinking preemption is enabled because trace_hardirqs_off() has not yet been called. Switch the order of these two calls, which matches their ordering in interrupt_enter_prepare. Fixes: 5f0b6ac3 ("powerpc/64/syscall: Reconcile interrupts") Signed-off-by:
-
Madhavan Srinivasan authored
Introduce code to support the checking of attr.config* for values which are reserved for a given platform. Performance Monitoring Unit (PMU) configuration registers have fields that are reserved and some specific values for bit fields are reserved. For ex., MMCRA[61:62] is Random Sampling Mode (SM) and value of 0b11 for this field is reserved. Writing non-zero or invalid values in these fields will have unknown behaviours. Patch adds a generic call-back function "check_attr_config" in "struct power_pmu", to be called in event_init to check for attr.config* values for a given platform. Signed-off-by:
Madhavan Srinivasan <maddy@linux.ibm.com> Signed-off-by:
-
Pu Lehui authored
Fix sparse warnings: arch/powerpc/platforms/pseries/rtas-fadump.c:250:6: warning: symbol 'rtas_fadump_set_regval' was not declared. Should it be static? Signed-off-by:
Pu Lehui <pulehui@huawei.com> Signed-off-by:
-
Christophe Leroy authored
Flushing functions don't rely on preemption being disabled, so use kmap_local_page() instead of kmap_atomic(). Signed-off-by:
-
Christophe Leroy authored
flush_dcache_page() is only a few lines, it is worth inlining. ia64, csky, mips, openrisc and riscv have a similar flush_dcache_page() and inline it. On pmac32_defconfig, we get a small size reduction. On ppc64_defconfig, we get a very small size increase. In both case that's in the noise (less than 0.1%). text data bss dec hex filename 18991155 5934744 1497624f 26423523 19330e3 vmlinux64.before 18994829 5936732 1497624f 26429185 1934701 vmlinux64.after 9150963 2467502 184548 11803013 b41985 vmlinux32.before 9149689 2467302 184548 11801539 b413c3 vmlinux32.after Signed-off-by:
-
Christophe Leroy authored
'And' the given page address with PAGE_MASK to help GCC. With the patch: 00000024 <__flush_dcache_icache>: 24: 54 63 00 26 rlwinm r3,r3,0,0,19 28: 39 40 00 40 li r10,64 2c: 7c 69 1b 78 mr r9,r3 30: 7d 49 03 a6 mtctr r10 34: 7c 00 48 6c dcbst 0,r9 38: 39 29 00 20 addi r9,r9,32 3c: 7c 00 48 6c dcbst 0,r9 40: 39 29 00 20 addi r9,r9,32 44: 42 00 ff f0 bdnz 34 <__flush_dcache_icache+0x10> 48: 7c 00 04 ac hwsync 4c: 39 20 00 40 li r9,64 50: 7d 29 03 a6 mtctr r9 54: 7c 00 1f ac icbi 0,r3 58: 38 63 00 20 addi r3,r3,32 5c: 7c 00 1f ac icbi 0,r3 60: 38 63 00 20 addi r3,r3,32 64: 42 00 ff f0 bdnz 54 <__flush_dcache_icache+0x30> 68: 7c 00 04 ac hwsync 6c: 4c 00 01 2c isync 70: 4e 80 00 20 blr Without the patch: 00000024 <__flush_dcache_icache>: 24: 54 6a 00 34 rlwinm r10,r3,0,0,26 28: 39 23 10 1f addi r9,r3,4127 2c: 7d 2a 48 50 subf r9,r10,r9 30: 55 29 d9 7f rlwinm. r9,r9,27,5,31 34: 41 82 00 94 beq c8 <__flush_dcache_icache+0xa4> 38: 71 28 00 01 andi. r8,r9,1 3c: 38 c9 ff ff addi r6,r9,-1 40: 7d 48 53 78 mr r8,r10 44: 7d 27 4b 78 mr r7,r9 48: 40 82 00 6c bne b4 <__flush_dcache_icache+0x90> 4c: 54 e7 f8 7e rlwinm r7,r7,31,1,31 50: 7c e9 03 a6 mtctr r7 54: 7c 00 40 6c dcbst 0,r8 58: 39 08 00 20 addi r8,r8,32 5c: 7c 00 40 6c dcbst 0,r8 60: 39 08 00 20 addi r8,r8,32 64: 42 00 ff f0 bdnz 54 <__flush_dcache_icache+0x30> 68: 7c 00 04 ac hwsync 6c: 71 28 00 01 andi. r8,r9,1 70: 39 09 ff ff addi r8,r9,-1 74: 40 82 00 2c bne a0 <__flush_dcache_icache+0x7c> 78: 55 29 f8 7e rlwinm r9,r9,31,1,31 7c: 7d 29 03 a6 mtctr r9 80: 7c 00 57 ac icbi 0,r10 84: 39 4a 00 20 addi r10,r10,32 88: 7c 00 57 ac icbi 0,r10 8c: 39 4a 00 20 addi r10,r10,32 90: 42 00 ff f0 bdnz 80 <__flush_dcache_icache+0x5c> 94: 7c 00 04 ac hwsync 98: 4c 00 01 2c isync 9c: 4e 80 00 20 blr a0: 7c 00 57 ac icbi 0,r10 a4: 2c 08 00 00 cmpwi r8,0 a8: 39 4a 00 20 addi r10,r10,32 ac: 40 82 ff cc bne 78 <__flush_dcache_icache+0x54> b0: 4b ff ff e4 b 94 <__flush_dcache_icache+0x70> b4: 7c 00 50 6c dcbst 0,r10 b8: 2c 06 00 00 cmpwi r6,0 bc: 39 0a 00 20 addi r8,r10,32 c0: 40 82 ff 8c bne 4c <__flush_dcache_icache+0x28> c4: 4b ff ff a4 b 68 <__flush_dcache_icache+0x44> c8: 7c 00 04 ac hwsync cc: 7c 00 04 ac hwsync d0: 4c 00 01 2c isync d4: 4e 80 00 20 blr Signed-off-by:
-
Christophe Leroy authored
__flush_dcache_icache() is usable for non HIGHMEM pages on every platform. It is only for HIGHMEM pages that BOOKE needs kmap() and BOOK3S needs flush_dcache_icache_phys(). So make flush_dcache_icache_phys() dependent on CONFIG_HIGHMEM and call it only when it is a HIGHMEM page. We could make flush_dcache_icache_phys() available at all time, but as it is declared NOKPROBE_SYMBOL(), GCC doesn't optimise it out when it is not used. So define a stub for !CONFIG_HIGHMEM in order to remove the #ifdef in flush_dcache_icache_page() and use IS_ENABLED() instead. Signed-off-by:
-
Christophe Leroy authored
flush_dcache_icache_hugepage() is a static function, with only one caller. That caller calls it when PageCompound() is true, so bugging on !PageCompound() is useless if we can trust the compiler a little. Remove the BUG_ON(!PageCompound()). The number of elements of a page won't change over time, but GCC doesn't know about it, so it gets the value at every iteration. To avoid that, call compound_nr() outside the loop and save it in a local variable. Whether the page is a HIGHMEM page or not doesn't change over time. But GCC doesn't know it so it does the test on every iteration. Do the test outside the loop. When the page is not a HIGHMEM page, page_address() will fallback on lowmem_page_address(), so call lowmem_page_address() directly and don't suffer the call to page_address() on every iteration. Signed-off-by:
-
Christophe Leroy authored
flush_coherent_icache() doesn't need the address anymore, so it can be called immediately when entering the public functions and doesn't need to be disseminated among lower level functions. And use page_to_phys() instead of open coding the calculation of phys address to call flush_dcache_icache_phys(). Signed-off-by:
-
Christophe Leroy authored
flush_coherent_icache() can use any valid address as mentionned by the comment. Use PAGE_OFFSET as base address. This allows removing the user access stuff. Signed-off-by:
-
Christophe Leroy authored
__flush_dcache_icache() is only used in mem.c. Move it before the functions that use it and declare it static. And also fix the name of the parameter in the comment. Signed-off-by:
-
Christophe Leroy authored
Cache flushing functions are in the middle of completely unrelated stuff in mm/mem.c Create a dedicated mm/cacheflush.c for those functions. Also cleanup the list of included headers. Signed-off-by:
-
Bixuan Cui authored
The sparse tool complains as follows: arch/powerpc/platforms/powernv/opal-core.c:74:16: warning: symbol 'mpipl_kobj' was not declared. This symbol is not used outside of opal-core.c, so marks it static. Reported-by:
Hulk Robot <hulkci@huawei.com> Signed-off-by:
Bixuan Cui <cuibixuan@huawei.com> Signed-off-by:
-
Pu Lehui authored
Fix sparse warning: arch/powerpc/xmon/xmon.c:4216:1: warning: symbol 'spu_inst_dump' was not declared. Should it be static? This symbol is not used outside of xmon.c, so make it static. Signed-off-by:
-
Bixuan Cui authored
The sparse tool complains as follows: arch/powerpc/perf/hv-24x7.c:229:1: warning: symbol '__pcpu_scope_hv_24x7_txn_flags' was not declared. Should it be static? arch/powerpc/perf/hv-24x7.c:230:1: warning: symbol '__pcpu_scope_hv_24x7_txn_err' was not declared. Should it be static? arch/powerpc/perf/hv-24x7.c:236:1: warning: symbol '__pcpu_scope_hv_24x7_hw' was not declared. Should it be static? arch/powerpc/perf/hv-24x7.c:244:1: warning: symbol '__pcpu_scope_hv_24x7_reqb' was not declared. Should it be static? arch/powerpc/perf/hv-24x7.c:245:1: warning: symbol '__pcpu_scope_hv_24x7_resb' was not declared. Should it be static? This symbol is not used outside of hv-24x7.c, so this commit marks it static. Reported-by:
-
Bixuan Cui authored
The sparse tool complains as follows: arch/powerpc/perf/isa207-common.c:24:18: warning: symbol 'isa207_pmu_format_attr' was not declared. Should it be static? This symbol is not used outside of isa207-common.c, so this commit marks it static. Reported-by:
-
Bixuan Cui authored
The sparse tool complains as follows: arch/powerpc/platforms/pseries/pmem.c:142:27: warning: symbol 'drc_pmem_match' was not declared. Should it be static? This symbol is not used outside of pmem.c, so this commit marks it static. Reported-by:
-
